AI mannequin gives a hypoglycemia early warning system when driving

Low blood sugar ranges (hypoglycemia) are one of the harmful issues of diabetes and pose excessive danger throughout cognitively demanding duties requiring advanced motor expertise, reminiscent of driving a automobile. The utility of present instruments to detect hypoglycemia is restricted by diagnostic delay, invasiveness, low availability, and excessive prices.

A current research printed within the journal NEJM AI gives a novel technique to detect hypoglycemia throughout driving. The analysis was the work of LMU scientists in collaboration with colleagues from the College Hospital of Bern (Inselspital), ETH Zurich, and the College of St. Gallen.

Of their research, the researchers collected knowledge from 30 diabetics as they drove an actual automobile. For every affected person, knowledge was recorded as soon as throughout a state with regular blood sugar ranges and as soon as throughout a hypoglycemic state. To this finish, every affected person was intentionally put right into a hypoglycemic state by medical professionals current within the automobile. The collected knowledge comprised driving alerts reminiscent of automobile pace and head/gaze movement knowledge—for instance, the pace of eye actions.

Subsequently, the scientists developed a novel machine studying (ML) mannequin able to routinely and reliably detecting hypoglycemic episodes utilizing solely routinely collected driving knowledge and head/gaze movement knowledge.

“This expertise might function an early warning system in vehicles and allow drivers to take vital precautions earlier than hypoglycemic signs impair their means to drive safely,” says Simon Schallmoser, doctoral candidate on the Institute of AI in Administration at LMU and one of many contributing researchers.

The newly developed ML mannequin additionally carried out nicely when solely head/gaze movement knowledge was used, which is essential for future self-driving vehicles. Professor Stefan Feuerriegel, head of the Institute of AI in Administration and challenge accomplice, explains, “This research not solely showcases the potential for AI to enhance particular person well being outcomes but additionally its position in enhancing security on public roads.”